MKL FATAL ERROR on Google Colab with mkl==2019.0

When using control version 0.9.1 on Google Colab and mkl version is 2019.0 the session crashes when running ts2ss() with the following runtime log:

Steps to recreate:
!pip install control
!pip install slycot (likely not related)
!pip show mkl (verify mkl version, my sessions have all been 2019.0 so far)

import control as ct
ct.tf2ss(ct.tf([1],[1,1]))

Session crashes with previous error in runtime logs.

The fix I have found is to update the mkl version to 2021.4.0. I am not familiar enough with the version changes nor have I tried other mkl versions, so I cannot say what other versions work.
!pip install mkl==2021.4.0

Probably should add a requirement on mkl version when installing control in Colab.

I have not tried to reproduce this on a local Jupyter notebook.

Hi,

python-control does not call any BLAS/LAPACK library routine directly. We don't have and won't add any requirement on MKL, OpenBLAS or any other implemetation of BLAS/LAPACK directly to python-control.

This is most likely an issue with the MKL/NumPy/Slycot combination of your instance in Google Colab. And as you have found out, it seems to be fixed by updating to a more recent version. I am not familiar with the innards of Colab, but it seems that !pip install slycot is the culprit here. On a regular Python environment this would download the source and compile the Fortran routines with the headers present on the system.

BTW, numpy on Colab says, that is not using MKL at all:

import numpy as np
np.show_config()``

blas_mkl_info:
  NOT AVAILABLE
blis_info:
  NOT AVAILABLE
openblas_info:
    libraries = ['openblas', 'openblas']
    library_dirs = ['/usr/local/lib']
    language = c
    define_macros = [('HAVE_CBLAS', None)]
    runtime_library_dirs = ['/usr/local/lib']
blas_opt_info:
    libraries = ['openblas', 'openblas']
    library_dirs = ['/usr/local/lib']
    language = c
    define_macros = [('HAVE_CBLAS', None)]
    runtime_library_dirs = ['/usr/local/lib']
lapack_mkl_info:
  NOT AVAILABLE
openblas_lapack_info:
    libraries = ['openblas', 'openblas']
    library_dirs = ['/usr/local/lib']
    language = c
    define_macros = [('HAVE_CBLAS', None)]
    runtime_library_dirs = ['/usr/local/lib']
lapack_opt_info:
    libraries = ['openblas', 'openblas']
    library_dirs = ['/usr/local/lib']
    language = c
    define_macros = [('HAVE_CBLAS', None)]
    runtime_library_dirs = ['/usr/local/lib']
Supported SIMD extensions in this NumPy install:
    baseline = SSE,SSE2,SSE3
    found = SSSE3,SSE41,POPCNT,SSE42,AVX,F16C,FMA3,AVX2
    not found = AVX512F,AVX512CD,AVX512_KNL,AVX512_KNM,AVX512_SKX,AVX512_CLX,AVX512_CNL,AVX512_ICL

If you compile Slycot with -DBLA_VENDOR=Generic, then you won't have to update MKL:

%env SKBUILD_CONFIGURE_OPTIONS="-DBLA_VENDOR=Generic"
!pip install slycot control
import control as ct

env: SKBUILD_CONFIGURE_OPTIONS="-DBLA_VENDOR=Generic"
Collecting slycot
  Downloading slycot-0.4.0.0.tar.gz (1.5 MB)
     |████████████████████████████████| 1.5 MB 8.8 MB/s 
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
    Preparing wheel metadata ... done
Collecting control
  Downloading control-0.9.1.tar.gz (357 kB)
     |████████████████████████████████| 357 kB 39.9 MB/s 
Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from slycot) (1.21.6)
Requirement already satisfied: scipy in /usr/local/lib/python3.7/dist-packages (from control) (1.4.1)
Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from control) (3.2.2)
Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->control) (3.0.8)
Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->control) (0.11.0)
Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->control) (2.8.2)
Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->control) (1.4.2)
Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from kiwisolver>=1.0.1->matplotlib->control) (4.2.0)
Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.1->matplotlib->control) (1.15.0)
Building wheels for collected packages: slycot, control
  Building wheel for slycot (PEP 517) ... done
  Created wheel for slycot: filename=slycot-0.4.0-cp37-cp37m-linux_x86_64.whl size=1418275 sha256=99f50c169426a9233c293bd29c813c51e4d9abb480661c45964ae2885fe47a2b
  Stored in directory: /root/.cache/pip/wheels/fa/b2/19/1e6568896fbabaf2a03d2fb6575211d281afcb9e00ce026cf6
  Building wheel for control (setup.py) ... done
  Created wheel for control: filename=control-0.9.1-py2.py3-none-any.whl size=364703 sha256=294faf3306345fc5a496a380203c8378d79344075a7b0ae67358db5f613cbea0
  Stored in directory: /root/.cache/pip/wheels/b1/50/22/8005107713828997d1af664fbd629c7f935a1b87fc2f4ba6c1
Successfully built slycot control
Installing collected packages: slycot, control
Successfully installed control-0.9.1 slycot-0.4.0

ct.tf2ss(ct.tf([1],[1,1]))

\[
\left(\begin{array}{rll|rll}
-1\phantom{.}&\hspace{-1em}&\hspace{-1em}\phantom{\cdot}&1\phantom{.}&\hspace{-1em}&\hspace{-1em}\phantom{\cdot}\\
\hline
1\phantom{.}&\hspace{-1em}&\hspace{-1em}\phantom{\cdot}&0\phantom{.}&\hspace{-1em}&\hspace{-1em}\phantom{\cdot}\\
\end{array}\right)
\]

Now you have to figure out how to get Colab to render LaTeX output like a real Jupyter Notebook or ignore it and use the regular __repr__()

@bnavigator

Hi @bnavigator / Ben!

I saw your comments here and decided to get in touch. I am also trying to run a very simple code on Google Colab using the Control + slycot library, but the Kernel restarts (crashes).

Were you able to solve the problem?

Best regards,

Nicolas Spogis

Colab Errors:

WARNING:root:kernel f5072b3f-42c0-4442-bb37-dd67744b63be restarted
KernelRestarter: restarting kernel (1/5), keep random ports
Intel MKL FATAL ERROR: Cannot load libmkl_avx2.so or libmkl_def.so.
INTEL MKL ERROR: /lib/x86_64-linux-gnu/libmkl_avx2.so: undefined symbol: mkl_sparse_optimize_bsr_trsm_i8.

Code on Google Colab:
!apt-get install gfortran cmake --fix-missing
!apt-get install libblas-dev liblapack-dev

import os
os.environ['BLA_VENDOR']="Generic"
!echo Using $$BLA_VENDOR BLAS implementation

!pip install -v slycot

!pip install control

import control
import slycot
import matplotlib.pyplot as plt

Define transfer function of the system

s = control.tf('s')

K1 = 5.0
T1 = 5.0
P1 = K1/(T1*s+1)

K2 = 10.0
T2 = 20.0
P2 = K2/(T2*s+1)

K3 = 5.0
T3 = 5.0
P3 = K3/(T3*s+1)

K4 = 5.0
T4 = 5.0
P4 = K4/(T4*s+1)

G = control.append(P1, P2, P3, P4) #Google Colab Crash HERE!

Check python-control/Slycot#183 (comment)

Now not crash, but i have this error:

AttributeError                            Traceback (most recent call last)
[<ipython-input-11-beb8f34beafd>](https://localhost:8080/#) in <cell line: 1>()
----> 1 G = control.append(P1, P2, P3, P4)

2 frames
[/usr/local/lib/python3.10/dist-packages/slycot/transform.py](https://localhost:8080/#) in td04ad(rowcol, m, p, index, dcoeff, ucoeff, tol, ldwork)
    622                                            *(dcoeff.shape + expectedshape)),
    623                                        -5)
--> 624         out = _wrapper.td04ad_c(m,p,index,dcoeff,ucoeff,n,tol,ldwork)
    625     else:
    626         raise SlycotParameterError("Parameter rowcol had an illegal value", -1)

AttributeError: module 'slycot._wrapper' has no attribute 'td04ad_c'

This is my Google Colab: https://colab.research.google.com/drive/1hb9XDIVeMirQ2gevnaU282yFTs4-KS_t?usp=sharing

I'm already getting a numpy error while importing control when running your colab.

My own colab notebook still works, including your code:

https://colab.research.google.com/drive/1kKaK3jI_omOTrLCLQGzSs1uTowN5Cx_v?usp=sharing

@bnavigator

Hi @bnavigator / Ben!

Thank you very very very much!

You help me a lot.

As "The Cure" said! It's Friday, I'm in love!

Best regards,

Nicolas

Navigation Menu

Repository navigation

MKL FATAL ERROR on Google Colab with mkl==2019.0 #730

Define transfer function of the system

Now not crash, but i have this error:

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

MKL FATAL ERROR on Google Colab with mkl==2019.0 #730

Description

Activity

bnavigator commented on May 2, 2022

bnavigator commented on May 2, 2022

Spogis commented on Feb 1, 2024

Define transfer function of the system

bnavigator commented on Feb 1, 2024

Spogis commented on Feb 1, 2024

Now not crash, but i have this error:

bnavigator commented on Feb 2, 2024

bnavigator commented on Feb 2, 2024

Spogis commented on Feb 2, 2024

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions